Characteristics of the gut microbiota in gestational diabetes mellitus associated with poor dietary habits: An observational study

Abstract

Gestational diabetes mellitus (GDM) is increasingly associated with alterations in gut microbiota, particularly under the influence of poor dietary habits during pregnancy. However, specific microbial characteristics associated with GDM remain unclear. To address this, we used 16S rRNA gene sequencing to compare the gut microbiota of pregnant women with GDM and healthy pregnant individuals. We conducted amplicon sequence variant analysis, assessed α- and β-diversity, performed taxonomic profiling, and applied linear discriminant analysis effect size (LEfSe) as well as random forest modeling to identify differences in microbial composition and potential diagnostic biomarkers. Although overall microbial diversity did not differ significantly between the groups, distinct taxonomic shifts were observed. These differences suggest the presence of microbial dysbiosis associated with GDM and poor dietary patterns. Enrichment of beneficial genera in healthy pregnant individuals, along with an increased abundance of potentially dysbiotic taxa in the GDM group, highlights the potential role of gut microbiota in GDM pathogenesis. Notably, Butyricicoccus and Lachnospiraceae_UCG-004 emerged as candidate biomarkers for GDM diagnosis or risk stratification. These findings provide a foundation for future investigations into microbiota-targeted interventions and the development of noninvasive diagnostic strategies for GDM.

1. Introduction

Gestational diabetes mellitus (GDM) is a form of hyperglycemia that develops during pregnancy, typically in the second or third trimester, and is characterized by elevated blood glucose levels.^[1,2] This condition arises when the maternal pancreas fails to produce adequate insulin to meet the increased metabolic demands of pregnancy.^[2] Unlike preexisting diabetes, GDM is generally transient and resolves after delivery.^[3] However, if inadequately managed, it poses substantial health risks to both the mother and fetus.^[4] The pathophysiology of GDM involves a combination of insulin resistance and impaired insulin secretion, influenced by hormonal fluctuations and placental factors specific to pregnancy.^[5] Despite advances in diagnosis and treatment, key aspects of GDM pathogenesis remain poorly understood.

Dietary habits play a pivotal role in the development and progression of GDM through multiple mechanisms. Excessive intake of refined carbohydrates, sugars, and saturated fats promotes insulin resistance, thereby increasing the risk of GDM.^[6] Conversely, insufficient consumption of fiber-rich foods, fruits, and vegetables impairs glycemic regulation and exacerbates metabolic dysfunction.^[7,8] Moreover, poor dietary habits have been associated with gut microbiota imbalance, characterized by reduced microbial diversity and a decline in beneficial taxa. This dysbiosis has been linked to altered metabolic states, further contributing to the development of GDM.^[9,10] Collectively, these mechanisms underscore the central role of nutrition in GDM pathogenesis and the importance of dietary interventions in its prevention and management.

The aim of this study was to identify gut microbiota-derived biomarkers of GDM in the context of poor dietary habits. First, we conducted a comprehensive assessment of dietary patterns and gut microbial composition in pregnant women diagnosed with GDM and in healthy pregnant individuals. Next, we utilized linear discriminant analysis effect size (LEfSe) to determine whether specific microbial taxa were differentially abundant in association with GDM. Finally, a Random Forest model, based on the area under the curve (AUC) metric, was applied to evaluate the predictive value of microbial features derived from 16S rRNA gene sequencing.

2. Methods

2.1. Collection of fresh fecal samples from volunteers

This study included 20 pregnant women aged 22 to 35 years, all at 37 to 41 weeks of gestation. The study was conducted between March 20, 2024 and April 23, 2024. Participants were required to have a prepregnancy body mass index within the normal range of 18.5 to 24.9 kg/m². All participants underwent a 75-g oral glucose tolerance test between the 24th and 28th weeks of gestation. The GDM group comprised 10 individuals who were diagnosed based on oral glucose tolerance test results meeting at least one of the following criteria: fasting plasma glucose (FPG) ≥92 mg/dL, 1-hour plasma glucose ≥180 mg/dL, or 2-hour plasma glucose ≥153 mg/dL. In contrast, the healthy pregnant (H) group included 10 individuals who did not meet any of these diagnostic thresholds. Exclusion criteria were as follows: smoking and/or alcohol consumption during pregnancy; use of antibiotics or probiotics within the past month; and diagnosis of inflammatory bowel disease, infectious diseases, cardiac disease, hepatic or renal disorders, immune system diseases, psychiatric disorders, or malignancies. All participants were residents of Hangzhou, Zhejiang Province. Written informed consent was obtained from all participants prior to enrollment. The study protocol was approved by the Ethics Review Committee of the Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (Approval No. 20240186). Fresh stool samples were collected and stored at 4°C immediately after defecation.

2.2. Quality of the dietary habits

The dietary habits of pregnant women were assessed using the validated index of diet quality questionnaire.^[11] A total score of <10 out of a maximum of 15 points was considered indicative of suboptimal dietary habits.

2.3. Genomic DNA extraction and 16S rRNA gene sequencing

Gut microbiota was assessed through 16S rRNA gene sequencing. Genomic DNA was extracted from the microbial communities using the FastDNA® Spin Kit for Soil (MP Biomedicals, Solon) in accordance with the manufacturer’s protocol. The bacterial 16S rRNA gene was amplified using a thermocycling polymerase chain reaction system (GeneAmp 9700, ABI, San Diego) with the primer pair 341F (5′-CCTAYGGGRBGCASCAG-3′) and 806R (5′-GGACTACHVGGTWTCTAAT-3′). polymerase chain reaction conditions consisted of an initial denaturation at 95°C for 3 minutes, followed by 27 cycles of denaturation at 95°C for 30 seconds, annealing at 55°C for 30 seconds, and extension at 72°C for 45 seconds. A final extension step was carried out at 72°C for 10 minutes. The amplified products were then purified and subjected to paired-end sequencing on the NovaSeq PE250 platform (Illumina, San Diego) at Majorbio Bio-Pharm Technology Co., Ltd., Shanghai, China.

After quality filtering and read assembly, the raw sequences were denoised using the DADA2 plug-in in QIIME2 (version 2020.2) to generate amplicon sequence variants. Taxonomic classification of amplicon sequence variants was performed using the SILVA 16S rRNA database (v138) and the Naïve Bayes classifier integrated in QIIME2. Alpha diversity was evaluated using Ace, Chao, and Sobs indices. Beta diversity was analyzed using the q2-diversity plug-in in QIIME2, with Bray–Curtis distances calculated to assess inter-sample variability. All sequencing data have been deposited in the National Center for Biotechnology Information Short Read Archive under accession number PRJNA1165346.

2.4. Data analysis

Statistical analyses were performed using SPSS version 23.0 (IBM Corporation, Chicago), and the results were presented as mean ± standard error of the mean (SEM). The Shapiro–Wilk test was used to assess the normality of data distribution. For data that conformed to a normal distribution, paired t-tests were applied. For non-normally distributed data, paired Wilcoxon rank-sum tests were used instead. Differences in α-diversity – measured using Ace, Chao, and Shannon indices – were evaluated via one-way analysis of variance (ANOVA), followed by Tukey–Kramer post hoc testing, based on the ASV table. To assess β-diversity, principal coordinate analysis (PCoA) and nonmetric multidimensional scaling (NMDS) were performed at the genus level using Bray–Curtis dissimilarity. Structural variation was further evaluated through permutational multivariate analysis of variance (PERMANOVA). To identify differentially abundant microbial taxa between H and GDM groups, linear discriminant analysis effect size (LEfSe) was employed. A threshold LDA score >2.0 was set to determine statistical significance. Random forest models were constructed for biomarker prediction, using base-10 logarithmic transformation for data normalization. Model performance was evaluated by calculating theAUC under the receiver operating characteristic curve (ROC). All data processing, statistical analysis, and graphical visualizations were conducted on the Majorbio Cloud Platform (https://www.majorbio.com).

3. Results

3.1. ASV analysis

Figure 1A and B present the pan and core analyses of gut microbiota composition in the H and GDM groups. The pan analysis reflects the total number of genera identified within each group, whereas the core analysis illustrates the number of genera shared between the 2 groups. In both analyses, the rate of increase or decrease in the total and shared genera gradually plateaued. This trend suggests that the sample sizes were adequate for capturing genus-level richness and identifying core microbial genera. As shown in Figure 1C, both the H and GDM groups exhibited high genus-level richness and an even distribution of gut microbiota. This was evidenced by the wide and gradually tapering range along the horizontal axis, indicating a balanced community structure with no single genus predominating.

Figure 1.

ASV analysis of gut microbiota. (A) Pan analysis and (B) core analysis between the H and GDM groups. (C) Rank-abundance curves between the H and GDM groups. ASV = amplicon sequence variant, GDM = gestational diabetes mellitus.

3.2. Diversity analysis

The diversity of the gut microbiota was assessed through 16S rRNA sequencing. The α-diversity indices – Ace (P = .273), Chao (P = .2413), and Shannon (P = .6232) – showed no statistically significant differences in community richness between the H and GDM groups (P > .05; Fig. 2A–C). Consistently, β-diversity analysis at the genus-level revealed no significant differences in microbial community structure between the 2 groups. This finding was supported by both PCoA (P = .915) and NMDS (P = .915), as shown in Figure 2D and E.

Figure 2.

Diversity analysis of gut microbiota. (A–C) The difference in α-diversity (Ace, Chao, and sobs indices) between the H and GDM groups. (D and E) The difference in β-diversity (the PCoA and NMDS analysis) between the H and GDM groups. Statistical differences: P < .05. GDM = gestational diabetes mellitus.

3.3. Composition and difference analysis

Figure 3A displays a bar plot representing the relative abundance of bacterial genera across 20 fecal samples. In the H group, Faecalibacterium was the most dominant genus, accounting for 15.72% of the total relative abundance. This was followed by Bacteroides (8.53%), Blautia (7.75%), and unclassified_k_norank_d_Bacteria (4.16%) (Fig. 3B). In contrast, the GDM group showed a different composition. Blautia was the most abundant genus, comprising 18.26% of the total, followed by Faecalibacterium (15.49%), Bacteroides (8.29%), and Bifidobacterium (4.85%) (Fig. 3C). To identify taxa with significantly different abundances between the 2 groups, LEfSe was performed. As shown in Figure 3D, a total of 10 genera exhibited significant differences (LDA score > 2.0). Among these, Butyricicoccus and Lachnospiraceae_UCG-004 were significantly enriched in the H group. In contrast, Fusicatenibacter, CAG-352, Moryella, and norank_f__Lachnospiraceae were more abundant in the GDM group.

Figure 3.

Composition and difference analysis of gut microbiota. (A) Community barplot analysis between the H and GDM groups on genus level. (B and C) Community pieplot analysis between the H and GDM groups on genus level. (D) LEfSe barplot analysis between the H and GDM groups on genus level. GDM = gestational diabetes mellitus, LEfSe = linear discriminant analysis effect size.

3.4. Biomarker identification

Furthermore, the potential utility of gut microbiota as a biomarker for GDM was evaluated using a random forest classification model. The AUC-based random forest algorithm was applied to identify the optimal model that maximized the AUC under ROC. In the validation cohorts comprising the H and GDM groups, the 8 most important genera were selected based on their feature importance scores to distinguish between the 2 groups (Fig. 4A). Subsequent specificity and sensitivity analyses of these top 8 genera yielded an AUC of 0.57 (95% CI: 0.29–0.85) (Fig. 4B).

Figure 4.

Biomarker identification from gut microbiota. (D) Bar plot showing the variable importance of gut microbiota at the genus-level constructed using random forest model analysis. (E) Performance of the model candidates assessed using the ROC analysis of gut microbiota at the genus level: AUC = <0.5, no diagnostic value; AUC = 0.5 to 0.7, low accuracy; AUC = 0.7 to 0.9, certain degree of accuracy; AUC = >0.9, high accuracy. AUC = area under the curve, ROC = receiver operating characteristic curve.

4. Discussion

The gut microbiota is a complex and dynamic ecosystem that plays a critical role in maintaining maternal health during pregnancy. In this study, we investigated how GDM and suboptimal dietary habits influence the composition of the gut microbiota. The interactions among GDM, dietary patterns, and microbial responses were analyzed using 16S rRNA gene sequencing.

The results shown in Figure 1A and B suggest that both the GDM and H groups harbor a well-characterized gut microbiota, exhibiting high diversity and even distribution. The gradual stabilization of total and shared genera further indicates that the sample sizes were sufficient to assess microbial richness and to identify core genera.^[12] This conclusion is corroborated by Figure 1C, which also demonstrates consistently high richness and evenness in both groups.^[13]

Moreover, the analysis revealed no significant differences in community richness or overall microbial structure between the H and GDM groups. This finding aligns with previous large-scale studies. For example, 1 study involving 1479 pregnant women employed 16S rRNA sequencing and similarly reported comparable gut microbiota profiles between GDM and non-GDM cohorts.^[14] Although suboptimal dietary habits are known to influence gut microbial composition,^[15,16] our results indicate distinct differences in microbial profiles between the H and GDM groups. This suggests that even in the absence of differences in overall richness or diversity, shifts at the taxonomic level may still occur. These findings emphasize the need to consider not only diversity indices but also taxonomic composition when evaluating gut microbiota alterations in metabolic disorders. Recent developments in metabolic medicine support dietary strategies, such as increasing fiber intake and consuming prebiotic-rich foods, as a way to modulate gut microbiota and improve metabolic health in individuals with GDM.^[17,18] In our study, Faecalibacterium was the most abundant genus in the H group. This taxon is widely recognized for its anti-inflammatory properties and its role in maintaining intestinal barrier integrity.^[19–21] In contrast, Blautia was more prevalent in the GDM group. However, the role of Blautia in metabolic diseases remains controversial, with studies reporting both beneficial and detrimental associations.^[22–24] These conflicting findings underscore the need for further investigation into the functional roles of specific taxa in the pathophysiology of GDM.

These conflicting findings underscore the need for further investigation into the functional roles of specific bacterial taxa in the pathophysiology of GDM. The identification of differentially abundant taxa between the GDM and H groups highlights a characteristic microbial dysbiosis associated with GDM. This observation aligns with growing evidence that gut microbiota dysregulation contributes to the development of metabolic disorders.^[25] The enrichment of Butyricicoccus and Lachnospiraceae_UCG-004 in the H group suggests a potential protective role against metabolic dysfunction and GDM.^[26–28] Conversely, the increased abundance of Fusicatenibacter, CAG-352, Moryella, and norank_f__Lachnospiraceae in the GDM group reflects a dysbiotic microbial profile potentially linked to metabolic disturbances.^[29,30]

To further explore these associations, a random forest classification model was constructed using the AUC-Random Forest algorithm. The model was optimized by incorporating the top 8 discriminatory genera. Among these, Lachnospiraceae_UCG-004 and Butyricicoccus emerged as potential key biomarkers for the diagnosis of GDM or for stratifying individuals according to metabolic risk. These findings contribute to ongoing discussions regarding the identification of reliable, accessible biomarkers for GDM.

This study has several limitations. The small sample size may reduce the statistical power and limit the generalizability of the findings. Although key differences in microbial composition were identified between groups, further validation in larger populations is needed. Additionally, the cross-sectional design prevents causal interpretation of the relationship between gut microbiota and GDM. Dietary data were not systematically collected, which may have introduced uncontrolled confounding. Finally, the use of 16S rRNA gene sequencing provides limited functional resolution. Future studies incorporating longitudinal sampling and multi-omics approaches are warranted.

5. Conclusion

This study offers novel insights into the relationship between gut microbiota composition and GDM, particularly as influenced by poor dietary habits. It advances our understanding of microbiota dysbiosis in the pathogenesis of GDM, underscores the role of dietary factors in shaping microbial communities, and supports the potential for microbiota-based biomarker discovery and personalized therapeutic interventions. Although no significant differences in overall gut microbial diversity were observed between pregnant women with and without GDM, specific bacterial taxa exhibited differential enrichment. This pattern is indicative of a subtle but biologically relevant dysbiosis associated with metabolic perturbations. The dominance of Faecalibacterium in the H group is consistent with its established role in maintaining gut homeostasis and anti-inflammatory activity. In contrast, the elevated presence of Blautia in the GDM group is noteworthy and warrants further investigation into its metabolic implications. Furthermore, the enrichment of Butyricicoccus and Lachnospiraceae_UCG-004 in the H group reinforces their potential utility as diagnostic or prognostic biomarkers for GDM. The application of the AUC-Random Forest algorithm to identify these genera underscores the utility of machine learning approaches in biomarker discovery and highlights the importance of noninvasive diagnostic strategies. This study contributes meaningful evidence to the complex interplay between gut microbiota, dietary behaviors, and GDM. It provides a foundation for future investigations into microbiota-targeted interventions and the development of personalized nutrition and diagnostic strategies for at-risk populations.

Acknowledgments

The authors would like to thank all the reviewers who participated in the review, as well as MJEditor (www.mjeditor.com) for providing English editing services during the preparation of this manuscript.

Author contributions

Conceptualization: Zhi Du, Lixia Bi.

Data curation: Linhua Hu, Hongli Liu, Fengbing Liang.

Formal analysis: Linhua Hu, Hongli Liu, Shudan Jiang.

Investigation: Guoxia Chen, Xiaoting Fang.

Methodology: Fengbing Liang, Zhi Du, Lixia Bi.

Project administration: Zhi Du, Lixia Bi.

Resources: Hongli Liu.

Software: Hongli Liu.

Supervision: Zhi Du, Lixia Bi.

Validation: Hongli Liu.

Writing – original draft: Linhua Hu, Hongli Liu, Fengbing Liang, Shudan Jiang, Guoxia Chen, Xiaoting Fang.

Writing – review & editing: Zhi Du, Lixia Bi.